GHS Urban Centre Database

Dot Map

Photo by Note Thanun on Unsplash

All urbanization, pushed beyond a certain point, automatically becomes suburbanization…
Every great city is just a collection of suburbs.
Its inhabitants do not live in their city; they merely inhabit it…
— Aldous Huxley

The database represents the global status on Urban Centres in 2015 by offering cities location, their extent (surface, shape), and describing each city with a set of geographical, socio-economic and environmental attributes. The database is built upon the Degree of Urbanisation, a definition used to outline the spatial extent of cities and settlements. The Urban Centre Database describes 13,135 Urban Centres.

Ingest

urban centres

df <- read.csv("archetypes/ghs-urban-centre-database/GHS_STAT_UCDB2015MT_GLOBE_R2019A_V1_2.csv", header = TRUE, stringsAsFactors = FALSE)
head(df, n = 10)

Analyze

create point clusters (using kmeans)

km <- kmeans(cbind(df$GCPNT_LAT, df$GCPNT_LON), centers = 30)
km_clusters <- data.frame("LATITUDE" = km$centers[,1], "LONGITUDE" = km$centers[,2], "SIZE" = km$size)
head(km_clusters, n = 10)

Analyze

create point clusters using dbscan

DBSCAN <- dbscan(cbind(df$GCPNT_LAT, df$GCPNT_LON), eps = 1.5, MinPts = 10)
coors <- data.frame("LONGITUDE" = df$GCPNT_LON, "LATITUDE" = df$GCPNT_LAT, "CLUSTER" = DBSCAN$cluster)
dist_clusters <- coors %>%
  group_by(CLUSTER) %>%
  summarise(
    LONGITUDE = mean(LONGITUDE, na.rm = T),
    LATITUDE = mean(LATITUDE, na.rm = T),
    SIZE = n()
  )
head(dist_clusters, n = 10)

Analyze

create point clusters using hierarchical clustering

hclust_clusters <- data.frame("LONGITUDE" = df$GCPNT_LON, "LATITUDE" = df$GCPNT_LAT)
hclust_clusters$CLUSTER <- cutree(hclust(dist(hclust_clusters[,1:2])), 30)
hclust_clusters <- hclust_clusters %>%
  group_by(CLUSTER) %>%
  summarise(
    LONGITUDE = mean(LONGITUDE, na.rm = T),
    LATITUDE = mean(LATITUDE, na.rm = T),
    SIZE = n()
  )
head(hclust_clusters, n = 10)

Base map

natural earth

ne_world <- ne_countries(scale = "small", returnclass = "sf")
ne_world <- filter(ne_world, iso_a3 != "ATA")

View

kmeans point clusters by size (count of members)

theme_opts <- theme(
  text = element_text(family = "inconsolata"), 
  plot.title = element_text(color = "black", size = 14, face = "bold", family = "inconsolata"),
  plot.subtitle = element_text(color = "black", size = 12, family = "inconsolata"),
  plot.caption = element_text(color = "#555555", size = 8, family = "inconsolata"),
  plot.background = element_blank(),
  panel.grid = element_blank(),
  panel.grid.major = element_blank(),
  panel.grid.minor = element_blank(),
  panel.background=element_rect(fill="white", colour="white"),
  panel.border = element_blank(),
  axis.text = element_blank(),
  axis.line = element_blank(),
  axis.ticks = element_blank(),
  axis.title = element_blank(),
  legend.position = "none",
  legend.title = element_blank()
)

color_palette <- paletteer_d("MapPalettes::bruiser")

v1 <- ggplot(data = ne_world) +
  geom_sf(fill="white", color="#888888", stroke=0.5) +
  geom_point( data = km_clusters, aes(x=LONGITUDE, y=LATITUDE, color=SIZE, size=SIZE), alpha = 1.0 ) +
  geom_text( data = km_clusters, aes(x=LONGITUDE, y=LATITUDE, label=SIZE), color = "white", vjust = 0.5, hjust = 0.5) +
  scale_colour_stepsn(colours = color_palette) +
  scale_size_continuous(range = c(6, 15)) +
  labs(x=NULL,
       y=NULL,
       title = "GHS Urban Centre Database", 
       subtitle="kmeans point clustering",
       caption="points are at cluster centers; size by cluster point members") +
  theme_bw() +
  theme_opts

girafe(ggobj = v1, width_svg = 16, height_svg = 7,
       options = list(opts_sizing(rescale = TRUE, width = 1.0)))

View

dbscan point clusters by size (count of members)

v2 <- ggplot(data = ne_world) +
  geom_sf(fill="white", color="#888888", stroke=0.5) +
  geom_point( data = dist_clusters, aes(x=LONGITUDE, y=LATITUDE, color=SIZE, size=SIZE), alpha = 1.0 ) +
  geom_text( data = dist_clusters, aes(x=LONGITUDE, y=LATITUDE, label=SIZE), color = "white", vjust = 0.5, hjust = 0.5) +
  scale_colour_stepsn(colours = color_palette) +
  scale_size_continuous(range = c(6, 20)) +
  labs(x=NULL,
       y=NULL,
       title = "GHS Urban Centre Database", 
       subtitle="dbscan point clustering") +
  theme_bw() +
  theme_opts

girafe(ggobj = v2, width_svg = 16, height_svg = 7,
       options = list(opts_sizing(rescale = TRUE, width = 1.0)))

View

hclust point clusters by size (count of members)

v3 <- ggplot(data = ne_world) +
  geom_sf(fill="white", color="#888888", stroke=0.5) +
  geom_point( data = hclust_clusters, aes(x=LONGITUDE, y=LATITUDE, color=SIZE, size=SIZE), alpha = 1.0 ) +
  geom_text( data = hclust_clusters, aes(x=LONGITUDE, y=LATITUDE, label=SIZE), color = "white", vjust = 0.5, hjust = 0.5) +
  scale_colour_stepsn(colours = color_palette) +
  scale_size_continuous(range = c(8, 20)) +
  labs(x=NULL,
       y=NULL,
       title = "GHS Urban Centre Database", 
       subtitle="hclust point clustering") +
  theme_bw() +
  theme_opts

girafe(ggobj = v3, width_svg = 16, height_svg = 7,
       options = list(opts_sizing(rescale = TRUE, width = 1.0)))

References

Citations and data sources

Florczyk, A.J., Corbane, C., Schiavina, M., Pesaresi, M., Maffenini, L., Melchiorri, M., Politis, 
P., Sabo, F., Freire, S., Ehrlich, D., Kemper, T., Tommasi, P., Airaghi, D. and L. Zanchetta. 2019. 
GHS Urban Centre Database 2015, multitemporal and multidimensional attributes, R2019A. 
European Commission, Joint Research Centre (JRC) [Dataset] 
PID: http://data.europa.eu/89h/53473144-b88c-44bc-b4a3-4583ed1f547e